This report is a data analysis investigation into racial disparity in deaths from COVID-19 in the United States, and the effects of certain health and demographic factors.
NOTE: Throughout the report the population is separated into two groups, “non-Hispanic White only”, and “not Non-Hispanic White only”, based on Census categories. For lack of a more feasible racially-sensitive construct, these will be referred to as “White” and “non-White”, respectively. Any analyses including “Hispanic Whites” in “White”, or not limiting to “White only”, or using any other racial groupings, could reach different, equally-valid conclusions.
There is racial disparity in COVID deaths in America
Any racial disparity will not be affected by age group
The sex and/or party of a State’s Governor has an effect on racial disparity in COVID deaths.
There are differences in health factors and outcomes by racial group
Some health factors contribute to COVID deaths
There are differences in population density between racial groups
Per capita COVID deaths scale with measures of population density
The data on population and racial breakdowns by states come from the Census. The Annual State Resident Population Estimates for 6 Race Groups (5 Race Alone Groups and Two or More Races) by Age, Sex, and Hispanic Origin document from 2020 gives the estimates as of 2020-07-01, the most recent publicly available figures.
The data on COVID deaths by race, state, and age, come from the Provisional Death Counts for Coronavirus Disease (COVID-19): Weekly State-Specific Data Updates, which is produced by the National Center for Health Statistics division of the CDC, using data from the National Vital Statistics System.
Data is absent for Vermont, Montana, Wyoming, Alaska, and Hawaii.
Figure 1: Racial Map of the United States
Figure 1 shows the concentration of non-White populations as a percentage of total population. Populations of note include: the Black population in the South and Chicago and Detroit, the large Hispanic population in the Southwest and Florida, and the noticeable populations of immigrants on the Pacific and North Atlantic coasts (and especially near New York City). The rest of the West, Midwest, and New England is noticeably lacking in non-White population except for in major urban areas.
Figure 2: Non-White COVID Deaths as proportion of total COVID Deaths in the United States
Figure 2 shows the percentage of COVID deaths in each state that occured within the non-White population. Ceteris paribus, this map should be roughly identical to figure 1. To meaningfully quantify the differences between figures 1 and 2, a new measure, the Racial Disparity Index (RDI) is introduced. The RDI is the ratio of “proportion of COVID deaths from among the non-White population” to “the proportion of the population that is non-White”. If deaths were equally distributed between racial groups, the RDI would be near 1. A value at or near zero indicates deaths occurred almost exclusively among the White population. A value of 2 indicates that twice as many deaths as expected occurred among the non-White population. The RDI is an imperfect ad hoc measure, but it can help give a simplified, intuitive sense of the racial disparity in COVID deaths.
Figure 3: A map of the Racial Disparity Index for the United States
As indicated in Figure 3, the RDI is low in a few states, somewhat elevated in the majority of states, and extremely high in some states.
Those states with the lowest RDIs (West Virginia, Maine, and New Hampshire) are the least diverse states in the nation. They are also among the states with the fewest overall deaths; none has more than a few hundred COVID deaths total. This combination of factors has led them to not experience a single reported non-White COVID death.
Those states with the highest RDIs include New York, Wisconsin, Utah, Michigan, and Missouri (see Table 1 and figure 8).
Figure 4: A Map of the Racial Disparity Index for Michigan
Figure 5: A Map of the Racial Disparity Index for New York
Among those worst states, the RDI seems to be worse in bigger cities like New York City, Detroit, Flint, and Grand Rapids (see figures 4 and 5).
The RDI varies somewhat by age bracket. The RDI for each age bracket was calculated for each Census Region.
Figure 6: Census Regions and Divisions of the United States
Figure 7 shows the differences and trends in age-based RDI for the different Census Regions.
NOTE: The figure does not average by state values, but across the entire region, so larger states are more influential.
Figure 7: RDI for each Census Region plotted by age bracket midpoint
The horizontal line at 1 is because there were no reported COVID deaths at all for that age group, not because of any actual racial parity. For all regions the general trend was a high, relatively stable RDI with a slight decrease toward parity with increasing age.
Figure 8: RDI for each Census Region by Age and State
Figure 8 shows the individual contributions of each state, with empty values (no COVID deaths for that state and age recorded in the dataset) removed for clarity.
Table 1: Sortable State Political Data and RDI by Age Bracket
| State | Governor Party | Governor Sex | 15-24 years | 25-34 years | 35-44 years | 45-54 years | 55-64 years | 65-74 years | 75-84 years |
|---|---|---|---|---|---|---|---|---|---|
| Alabama | Republican | Female | 1.00 | 1.00 | 2.72 | 2.43 | 2.12 | 2.18 | 1.81 |
| Alaska | Republican | Male | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Arizona | Republican | Male | 1.00 | 1.85 | 1.97 | 2.00 | 2.30 | 2.60 | 2.35 |
| Arkansas | Republican | Male | 1.00 | 1.00 | 1.00 | 1.00 | 3.10 | 1.60 | 1.95 |
| California | Democrat | Male | 1.38 | 1.48 | 1.41 | 1.41 | 1.57 | 1.64 | 1.50 |
| Colorado | Democrat | Male | 1.00 | 1.00 | 2.99 | 2.41 | 2.77 | 2.64 | 2.08 |
| Connecticut | Democrat | Male | 1.00 | 1.00 | 2.43 | 1.95 | 2.29 | 2.04 | 1.75 |
| Delaware | Democrat | Male | 1.00 | 1.00 | 1.00 | 1.21 | 1.62 | 1.34 | 1.12 |
| District of Columbia | Democrat | Female | 1.00 | 1.00 | 1.79 | 1.53 | 1.42 | 1.44 | 1.29 |
| Florida | Republican | Male | 1.00 | 1.85 | 1.59 | 1.61 | 1.75 | 1.98 | 1.73 |
| Georgia | Republican | Male | 1.00 | 1.90 | 1.93 | 1.76 | 1.76 | 1.87 | 1.53 |
| Hawaii | Democrat | Male | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Idaho | Republican | Male | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 0.00 | 0.00 |
| Illinois | Democrat | Male | 1.00 | 2.27 | 2.22 | 2.24 | 2.56 | 2.44 | 2.09 |
| Indiana | Republican | Male | 1.00 | 1.00 | 0.00 | 3.32 | 2.80 | 2.66 | 2.23 |
| Iowa | Republican | Female | 1.00 | 1.00 | 1.00 | 0.00 | 3.50 | 3.18 | 0.00 |
| Kansas | Democrat | Female | 1.00 | 1.00 | 1.00 | 1.00 | 3.58 | 3.58 | 2.09 |
| Kentucky | Democrat | Male | 1.00 | 1.00 | 1.00 | 0.00 | 2.25 | 2.61 | 1.66 |
| Louisiana | Democrat | Male | 1.00 | 2.21 | 2.37 | 2.23 | 2.05 | 2.02 | 1.75 |
| Maine | Democrat | Female | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 0.00 | 0.00 |
| Maryland | Republican | Male | 1.00 | 1.86 | 1.83 | 1.76 | 1.82 | 1.77 | 1.59 |
| Massachusetts | Republican | Male | 1.00 | 1.00 | 1.93 | 1.92 | 1.99 | 1.76 | 1.45 |
| Michigan | Democrat | Female | 1.00 | 2.04 | 2.57 | 2.99 | 3.44 | 3.65 | 3.02 |
| Minnesota | Democrat | Male | 1.00 | 1.00 | 1.00 | 5.71 | 4.85 | 4.20 | 1.78 |
| Mississippi | Republican | Male | 1.00 | 1.00 | 2.17 | 2.02 | 1.89 | 1.88 | 1.88 |
| Missouri | Republican | Male | 1.00 | 1.00 | 1.00 | 3.57 | 3.67 | 4.01 | 2.59 |
| Montana | Democrat | Male | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 0.00 |
| Nebraska | Republican | Male | 1.00 | 1.00 | 1.00 | 5.21 | 5.00 | 4.20 | 0.00 |
| Nevada | Democrat | Male | 1.00 | 1.00 | 1.00 | 1.46 | 2.00 | 1.59 | 1.79 |
| New Hampshire | Republican | Male | 1.00 | 1.00 | 1.00 | 1.00 | 0.00 | 0.00 | 0.00 |
| New Jersey | Democrat | Male | 1.00 | 1.55 | 1.56 | 1.79 | 1.88 | 1.77 | 1.54 |
| New Mexico | Democrat | Female | 1.00 | 1.45 | 1.49 | 1.60 | 1.92 | 1.95 | 1.81 |
| New York | Democrat | Male | 2.04 | 1.67 | 1.78 | 1.86 | 2.00 | 2.06 | 1.87 |
| North Carolina | Democrat | Male | 1.00 | 1.00 | 2.47 | 2.21 | 2.05 | 2.16 | 2.05 |
| North Dakota | Republican | Male | 1.00 | 1.00 | 1.00 | 1.00 | 0.00 | 0.00 | 0.00 |
| Ohio | Republican | Male | 1.00 | 1.00 | 1.77 | 1.99 | 2.27 | 2.26 | 1.63 |
| Oklahoma | Republican | Male | 1.00 | 1.00 | 1.00 | 1.00 | 1.48 | 1.64 | 0.59 |
| Oregon | Democrat | Female | 1.00 | 1.00 | 1.00 | 1.00 | 2.97 | 1.64 | 0.00 |
| Pennsylvania | Democrat | Male | 1.00 | 1.00 | 2.71 | 2.87 | 3.17 | 3.20 | 2.49 |
| Rhode Island | Democrat | Female | 1.00 | 1.00 | 1.00 | 1.00 | 2.12 | 1.86 | 1.21 |
| South Carolina | Republican | Male | 1.00 | 1.00 | 2.59 | 2.31 | 1.98 | 2.20 | 1.86 |
| South Dakota | Republican | Female | 1.00 | 1.00 | 1.00 | 1.00 | 0.00 | 0.00 | 0.00 |
| Tennessee | Republican | Male | 1.00 | 1.00 | 3.55 | 2.98 | 2.85 | 3.16 | 2.59 |
| Texas | Republican | Male | 1.50 | 1.37 | 1.47 | 1.50 | 1.69 | 1.76 | 1.62 |
| Utah | Republican | Male | 1.00 | 1.00 | 1.00 | 1.00 | 3.30 | 3.12 | 2.27 |
| Vermont | Republican | Male | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 0.00 | 0.00 |
| Virginia | Democrat | Male | 1.00 | 1.00 | 2.34 | 2.22 | 2.28 | 2.05 | 1.59 |
| Washington | Democrat | Male | 1.00 | 1.00 | 2.86 | 1.97 | 2.48 | 2.16 | 1.54 |
| West Virginia | Republican | Male | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 0.00 | 0.00 |
| Wisconsin | Democrat | Male | 1.00 | 1.00 | 1.00 | 4.12 | 5.95 | 5.37 | 5.18 |
| Wyoming | Republican | Male | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
Table 1 gives the RDI age breakdowns and Governor information for each state. Figures 9 and 10 show that information plotted by Governor sex and party, respectively. The female-led states with worse RDIs (Michigan, Kansas, Iowa) had higher populations, leading to the increased but not statistically-significant RDI averages (see table 2). For both male-led states and states of either party (regardless of Governor sex), the RDIs were approximately 2 for all ages with reported COVID deaths. As of yet, the party of the Governor appears to make no significant difference (see table 3) in racial disparity in COVID deaths. Whether this is surprising or not is left as an exercise to the reader.
Figure 9: RDI of States Plotted by Age and Sex of Governor
Figure 10: RDI of States Plotted by Age and Party of Governor
Table 2: Analysis of Variance of Sex of State Governor vs RDI
| Df | Sum Sq | Mean Sq | F value | Pr(>F) | |
|---|---|---|---|---|---|
Governor Sex
|
1 | 0.8623252 | 0.8623252 | 2.994188 | 0.1091722 |
| Residuals | 12 | 3.4559960 | 0.2879997 | NA | NA |
Table 3: Analysis of Variance of Party of State Governor vs RDI
| Df | Sum Sq | Mean Sq | F value | Pr(>F) | |
|---|---|---|---|---|---|
Governor Party
|
1 | 0.0010211 | 0.0010211 | 0.0437133 | 0.8378951 |
| Residuals | 12 | 0.2803190 | 0.0233599 | NA | NA |
Tables 2 and 3 show that neither the sex nor the party of a state’s governor makes a statistically significant contribution to racial disparity in COVID deaths.
Using data from the Behavioral Risk Factor Surveillance System (BRFSS), health factors and living conditions affecting the non-White population were analyzed. Figure 11 shows the significant negative correlation between pneumonia vaccination and per capita COVID deaths among the elderly non-White population. The data were similarly negatively-correlated for the elderly White population, but the effect was orders of magnitude weaker (data not shown). The noticeably lower rate of pneumonia vaccination among the non-White population (table 4) is then a potential source of the differential rate of death from COVID, especially among the elderly.
Table 4: Pneumonia Vaccination by Racial Group
| Racial Group | Percent Vaccinated Against Pneumonia |
|---|---|
| Non-White | 23.68 % |
| White | 37.29 % |
Figure 11: The correlation between pneumonia vaccination and per capita COVID deaths among the elderly non-White population
While statewide rates of pneumonia vaccination was significantly correlated with per capita COVID deaths, other health variables (see figure 12; many, many others not shown) did not reach statistical significance. While it is possible that this reflects there actual lack of correlation, the more likely explanation is the lack in granularity when examining state-level data. Varying levels of evidence exist for different conditions. Issues in reporting from different state- and county-level departments of health and coroners, and lack of availability of comorbidity data contribute to the difficulty in determining these correlations. Additionally, lack of COVID testing and other factors can lead to deaths not being correctly included in COVID counts.
Figure 12: The insignificant correlation between asthma and per capita COVID deaths among the elderly non-White population
Using data taken from the BRFSS, the effect of various measures of population density were analyzed. Using the United States Office of Management and Budget definition of Metropolitan Statistical Areas (MSAs, figure 13), the correlations between living in metropolitan areas, including (figure 14) or excluding (figure 15) the suburbs, and per capita COVID deaths among the non-White population were determined.
Figure 13: Metropolitan Statistical Areas
Figure 14: The correlations between living in MSAs and per capita COVID deaths among the non-White population
Figure 15: The correlations between living in MSAs (excluding suburbs) and per capita COVID deaths among the non-White population
Figure 16: The correlations between living in rural counties and per capita COVID deaths among the non-White population
Figure 17: The correlations between population and per capita COVID deaths among the non-White population
As shown in figures 14-17, all measures of density were significantly correlated with the severity of COVID pandemic. Statewide measures of the percentage of the population living in metropolitan areas showed strong positive relationships with per capita deaths from COVID among non-White populations of all ages. The expected opposite trend held for the statewide percentage of the population living in rural areas (figure 16). More directly, population density (figure 17) showed the strongest correlation with per capita COVID deaths across all age groups, races, and variables investigated (data not shown). These data are in line with the expectation that human-to-human diseases kill most readily in areas where humans are closer together.
Table 5: Metropolitan/Rural Residence by Racial Group
| Racial Group | Percent Living in MSAs | Percent Living in Rural Counties |
|---|---|---|
| Non-White | 92.04 % | 3.34 % |
| White | 80.87 % | 8.41 % |
The analyses revealed widespread, significant racial disparity in COVID deaths. The racial disparity was largely stable across states and age groups. No political factors were found to significantly contribute to racial disparity in COVID deaths. Noteable differences in health and demographic factors were uncovered, including pneumonia vaccination. While some health factors were found to contribute to deaths to COVID, without better reporting of death and comorbidity data, the evidence for the contributions of other health and demographic factors is too sparse for additional meaningful conclusions to be drawn yet. Most conclusively, the increased probability of living in more population-dense conditions among the non-White population (table 5) is likely to contribute seriously to difference in deaths. In short, close proximity to other humans is likely to be the strongest predictor of mortality from COVID, and a major driver of racial disparity in pandemic deaths.
Special thanks to the following sources of data: